From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alexander Duyck Subject: [net-next PATCH 0/2] Prevent packet reordering on tap/tun interfaces Date: Wed, 01 Jun 2016 18:17:11 -0700 Message-ID: <20160602010741.9319.56926.stgit@localhost.localdomain> Mime-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit To: netdev@vger.kernel.org, davem@davemloft.net, alexander.duyck@gmail.com Return-path: Received: from mail-pa0-f46.google.com ([209.85.220.46]:34602 "EHLO mail-pa0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751326AbcFBBRO (ORCPT ); Wed, 1 Jun 2016 21:17:14 -0400 Received: by mail-pa0-f46.google.com with SMTP id bz2so1253483pad.1 for ; Wed, 01 Jun 2016 18:17:13 -0700 (PDT) Sender: netdev-owner@vger.kernel.org List-ID: This patch set is meant to address a recent issue I found with VMs sending traffic across a network. Specifically what I found was that the tap interfaces were spreading single flows across all CPUs on the system due to the fact that the sending VM was being load balanced across them. Under light load this doesn't have much of an impact, however under heavier loads where the interfaces are already running with Rx cleanup from other flows on the CPUs I have seen this have a fairly significant impact as we can avoid any issues due to reordering. The results can be anywhere from 10 to 30% improvement from what I have seen. --- Alexander Duyck (2): net: Add function to allow configuration of RPS tun: Configure Rx queues to default to RPS enabled drivers/net/tun.c | 4 +++- include/linux/netdevice.h | 9 ++++++++ net/core/dev.c | 51 ++++++++++++++++++++++++++++++++++++++++++++- net/core/net-sysfs.c | 45 +++++++--------------------------------- 4 files changed, 70 insertions(+), 39 deletions(-)